View Full Version : How would you approach this photo?


oyr66
12-26-2004, 02:11 PM
Hello there,

Recently I got a bunch of old family photo circa 1910-1913. The posted pic is one of the worst. Being new to a photo restoration I would appreciate your input on what would be your steps here.

Thanks and Happy Holidays,

-oyr

chrishoggy
12-26-2004, 03:50 PM
Deleted and leaving this forum due to the actions of member rondon (http://www.retouchpro.com/forums/member.php?userid=366)

oyr66
12-26-2004, 05:38 PM
Chris,

Thanks for the advise. I do have an original. I scanned it at 600 dpi - the baby face is lost completely. So I would need to give this baby a new face and somehow blend it with the rest. I see it as a long project and hopefully will learn a thing or two as I go.

-oyr

Lin Evans
12-28-2004, 11:18 PM
I notice that in many of the faces one eye is good while the other is really bad. This seems strange, almost as if the original were defective. Also in some cases half of the face or shoulder is obliterated while the other half is decent.

Since humans are reasonably bi-symmetrical, it's possible to select the good half, copy and paste then move the paste, use PhotoShop's "Free Transform" tool to do a horizontal flip and manual rotate to position the selection so that the good side is duplicated as a replacement for the bad side.

Another technique where a feature is obliterated, such as a nose, is to find another person with similar color, tone and features and clone that missing feature.

On the badly distorted backdrop, select an area where the background is reasonably intact and use this as the "color" for the brush's picker. Simply paint the backdrop rather than attempt to use clone, healing brush, etc., to "fix" the really bad cracks. Once you have finished you can select the backdrop and apply some grain to simulate the original texture.

With the baby whose face is totally missing, you could select another child, match the size and simply clone the features. Unfortunately there is no way to accurately reproduce totally missing features.

Lin